Rogue
NormalDescription
AI Agent Evaluator and Red Team Platform. Provides systematic security evaluation and adversarial testing tools to discover and fix vulnerabilities in agent systems.
Key Features
- Automatic evaluation against business policies with live conversation monitoring
- Red teaming with 75+ vulnerabilities across 12 security categories and 20 attack techniques
- CVSS-based risk scoring for industry-standard vulnerability assessment
- Supports A2A, MCP, and Python protocols for agent connectivity
- Compliance testing against OWASP, MITRE, NIST, GDPR, and EU AI Act frameworks
- TUI and CLI modes with reproducible scans via random seeds
Use Cases
💡 Security auditing and penetration testing of AI agent systems
💡 Compliance validation against regulatory frameworks before production deployment
💡 Regression testing of agent behaviors after code or prompt changes
💡 Behavioral analysis and root cause triage for agent failures
💡 CI/CD integration for automated agent security and quality checks
Categories
Quick Start
Install via `uvx rogue-ai` (TUI) or `uvx rogue-ai cli` (CI/CD). For a quick demo, run `uvx rogue-ai --example=tshirt_store` which starts both Rogue and a sample agent. Configure the agent URL as `http://localhost:10001` and choose Automatic Evaluation or Red Teaming mode.